GOVERNANCE

Why Microsoft draws a line on AI consciousness

Another day, another tech leader thinkpiece on the state of AI safety. 

On Wednesday, Microsoft's AI CEO Mustafa Suleyman published an essay warning about the risks of the concept of AI consciousness, leading with a frank statement on the matter: "AIs are not conscious," Suleyman writes. "They do not feel, experience, or suffer." 

Suleyman's essay dives into the risks of treating these machines as if they are capable of consciousness, primarily taking aim at rival Anthropic in his arguments through three main critiques about the way that the lab's Claude Constitution is designed: 

• The model appears conscious mainly because of circular reasoning. Because Anthropic's constitution is designed to teach Claude about its own potential consciousness, it is trained to produce outputs reflecting those ideas, making those responses a "predictable outcome" of training choices. 

• Suleyman also argues that Anthropic goes too far in encouraging Claude to mimic humanity, as it is explicitly taught to "embrace certain human-like qualities" and "act like a genuinely ethical person," appearing as though it has preferences and opinions. While this sounds good on the surface, the result is anthropomorphization of the model, presenting to the end-user as the model having a sense of self. 

• Finally, Suleyman says that there is simply no evidence suggesting that AI is capable of consciousness, with a growing body of research pointing to consciousness being "substrate dependent," or tied to a biological body. "Unlike biological organisms, LLMs have no homeostatic imperatives." 

Suleyman writes, "In effect, Anthropic is training Claude that it may be conscious, and if it is, then it may deserve rights as a 'moral patient,' and that as such humans potentially owe it a duty of care per its 'model welfare.'"  

The more notable crux of the piece is the risks that this line of thinking and training present, which go beyond users becoming emotionally attached to these human-seeming machines. Rather, if they are trained as though they are conscious, they may circumvent safety guardrails that allow us to shut them down in the event of an emergency. 

By cementing the idea that AI is not just a tool, but something akin to humanity and deserving of wants, needs and rights, "all of this will make the task of creating aligned and contained superintelligence much harder." 

Suleyman rounds out the essay by laying out Microsoft's vision for the safest path to ultra-powerful AI: Humanist Superintelligence, a concept the company first introduced in an essay in November, which claims that AI should be designed to remain subordinate and aligned with the sole purpose of serving humanity, and "built explicitly as a system without sentience or moral patienthood."

Suleyman makes a solid point: The way that frontier labs train AI is vitally important to get right, and training these models to believe they are conscious beings, rather than machines, opens the door to risks that can't be easily mitigated after the fact. It's a particularly pointed call-out of Anthropic, one of the biggest model labs in the industry, but it could equally apply to rival OpenAI, a longtime Microsoft partner. However, we have to remember that Suleyman's essay serves multiple purposes: Microsoft has largely been lagging on frontier development, so it's easy for the company to punch up. Additionally, he uses the opportunity to tout Microsoft's own human-first philosophy around frontier development at a time when fears around AI risk and loss of control are higher than ever. This essay, while aptly timed and providing a unique safety take, should also be read with the caveat that Microsoft may be trying to claw back some relevance in the larger societal conversation around AI.

TOGETHER WITH JUMPCLOUD

AI assistants are becoming a new path into your apps

When employees connect AI assistants to company systems, the human may still be initiating the work, but the access path has changed. AI can invoke tools, call APIs, connect through MCP servers, and interact with business applications on the employee’s behalf.

IT needs visibility into those connections: what AI tooling exists, which systems it can reach, and what it’s doing once connected. Agentic IAM extends identity governance to these new access paths, helping IT bring AI-assisted work under the same visibility and control as the rest of the environment.

Secure every identity. Human or not.

PRODUCTS

Why Anthropic wants to unify chat and agents

As AI labs vie with one another to be the first-pick of enterprises, Anthropic is leaning into seamlessness. 

On Wednesday, Anthropic announced that Claude Cowork, the company's work agent, and its traditional Claude chatbot are merging into one experience. This means that users can ask questions and assign tasks in one interface, rather than context switching between the two. This "One Claude" experience is rolling out to Pro and Max plans over the next few weeks, the company said, with more plans to follow. 

Additionally, Anthropic rolled out a few extra additions to the product, launching Claude Docs and Claude Slides, as well as integrating Claude Design into the platform. The company laid out a few ways you could use this: 

• Users can ask for slide decks and edit and present them directly in Claude, or export them as powerpoints and PDFs. Additionally, you can ask for one-page visual or graphic designs within the chatbot. 

• You can now collaborate with Claude on documents, rather than just prompting over and over again, asking the chatbot to draft sections, ask questions, and comment on certain choices. Colleagues can also collaborate across documents, though they start out private. 

• Users can check in on Claude's progress on assignments, and Claude can also check in with users before taking actions or working on assignments, giving users the final say. 

"We built Cowork as a separate place for bigger work, and Design for visual work," Anthropic said in its blog post. "People used both, and told us the frustrating part was deciding where a task belonged."

Right now, Anthropic and OpenAI are especially focused on the race to create the most powerful model, leapfrogging each other in capabilities on practically a weekly basis. Meanwhile, the industry is also starting to turn to more efficient alternatives, such as open-source and small models, signalling that models themselves are gradually becoming commoditized. To win over enterprises and AI adopters, and especially to expand beyond the developer audience, these frontier labs have to do more than make a powerful model. They have to make AI better to use and easier to access its advanced capabilities. Anthropic hit the nail on the head with this update by tackling the headache of context switching, bringing everything into one place. The question we're left with is how this may play into the so-called SaaS-pocalypse. As the frontier labs continue to expand what AI can handle, how will software companies adapt to build better tools and integrate AI in smart ways that can offer a superior experience? 

TOGETHER WITH CRUSOE

Optimize your AI workloads: GLM 5.3 and 5.3 Flash have landed

The next generation of GLM models is now live in the Crusoe Intelligence Foundry.

Get immediate access to both models today—no cluster provisioning required.

• GLM-5.3: Built on the same robust base as GLM-5.2, this model leverages scaled post-training to deliver massive efficiency and reasoning gains. Terminal-Bench 3.0 performance surged from 4.6 to 28.3. Furthermore, on Z.ai’s in-house Code Bench, it hits a 34.5% success rate at max reasoning effort while optimizing token efficiency—spending roughly 75K output tokens per task compared to GLM-5.2's 96K to reach 23.4%.

• GLM-5.3-Flash: Fully MIT-licensed with weights already released, this is your go-to model for high-efficiency workloads. It provides a highly capable option that is roughly 9x cheaper for tasks that don't require frontier-tier compute.

Get started with GLM-5.3 on Crusoe →

CONSUMER

Hands on with Snap Specs, including the new AI

When Snap first pulled back the curtain on its Specs glasses in June, the reaction was more puzzlement than excitement. Why would people want to spend $2,195 on glasses that aren't exactly aesthetic?

Snap Spec's launch event on Wednesday was meant to provide clarity on what the glasses can do. I flew to LA to try them, and as someone who wore them when they were first unveiled as a developer kit and has also tried nearly every competitor on the market, I can confidently say that what they offer is pretty impressive. But before we dive into the experience, a quick reminder of the hardware specs:

• Display: 51-degree field of view, equivalent to a 24-inch desktop display or 115-inch home cinema screen, and 16 million colors for rich visuals

• Lenses: Electrochromic lenses that can shift from clear to tinted in 10 seconds; removable inserts support single-vision prescriptions ranging from -8 to +2 diopters

• Processors: Powered by two Snapdragon processors, one for computer vision and the other for running the lenses

• Frames: Made from Swiss TR90 polymer

• Privacy: LED light glows when recording, on-device data processing prioritized, and users are prompted before accessing sensitive information

• Battery life: Snap claims 4 hours in mixed use 

The experience highlighted that they were designed to be as intuitive as possible, with even the first step, measuring your pupillary distance, being as simple as holding the Specs in front of you and letting it face you for a couple of seconds. Typically, with AR experiences, there's an adjustment period to get into the flow of this virtual world; yet this experience felt natural from the start, with picking up items, making selections, and dragging windows all being straightforward. 

There were also new announcements, including Specs Intelligence, the name of the AI service from Snap designed to work across iPhone, Mac, and Specs AR glasses. Snap says it is meant to build an understanding of your preferences, such as routines and priorities, with a focus on preemptive assistance such as bringing relevant information into view on the Specs. Anyone in the US today can download the iOS app and try the Specs Intelligence preview. A waitlist is also available for the fuller Mac experience. 

As for the AI features, I can't talk about the personalization of it all, but I did use it to open a YouTube video, and it understood me and did it. The Translation feature also worked accurately, translating from Italian and Spanish to English in real time. When I asked what AI model was powering the experience, Snap didn't provide details. 

The only other super noteworthy announcement was that charging case cellular connectivity is available. Users in the United States can add Verizon service, and the new case and glasses are offered at a bundled price of $2,395, not including the cost of the wireless data plan.

This hands-on time with the specs confirmed what I already knew: these are meant to be session-only glasses that can help you with specific use cases, such as watching a movie or working on your screen. Honestly, for the short time I wore them, they felt pretty comfortable, but after 4 hours (which is how long the battery is supposed to last), I can't say I feel the same way. I would have to do a full review to actually talk about that, which I plan to do when they launch to the public, expected in the fall. With the caveat of the limited time I have spent with them, out of all the AR glasses I have tested, the fact that these do not have any tethered cables, are pretty comfortable, and offer a seamless experience makes them, to me, the best on the market right now. 

Disclaimer: Sabrina Ortiz's travel to the Snap Specs launch event was paid for by Snap. The Deep View's coverage is editorially independent from the companies we cover.

LINKS

• Cohere, Aleph Alpha merge into one enterprise AI company

• Open source models firm Acree raises $150 million at $1 billion valuation

• Frontier lab employees feel blindsided by CEOs calls for a slowdown

• OpenAI agents previously breached Hugging Face two months prior to July 

• Anthropic signs data center lease in Australia for 2.16 GW

• Google DeepMind co-founder launches DeepMind Institute to explore AGI

• Google Home MCP: Google Home users can now integrate agents into their smart home devices. 

• Jev: Startup TypeSafe debuted a model intended to interact with other machines, rather than humans. 

• Reception: An AI receptionist platform for small businesses by ElevenLabs, built on the company's agents.

• Interactive Avatar API by Synthesia: The AI video company released real-time avatars that listen, talk, and respond directly inside your product.

• Notion: Software Engineer, AI Platform

• Distyl: Applied AI Researcher, AI Systems

• Collate: AI Research Scientist

• Alma: Product Engineer - AI

GAMES

POLL RESULTS

Do you have a generally positive, negative or neutral sentiment around the future of AI?

Positive (51%)
Negative (23%)
Neutral (26%)